This article was born in response to an article . I do not agree with most of the author's theses, and I want to not only refute them in this article, but also put forward my own theses, which I will then defend in the comments. Next, I will talk about several principles that I adhere to when designing a local network for any enterprise.
The first principle is reliability. An unreliable network will always be more expensive due to the cost of its maintenance, losses during downtime and losses from outside interference. Based on this principle, I always design the main network only by wire, and, if necessary, additional wireless (guest network or network for mobile terminals). Why is the wireless network less reliable? Any wireless network has a number of security, stability, and compatibility issues. Too many risks for a serious company.
Reliability also determines the structure of the network. Topology "Star" is an ideal to strive for. Zvezda reduces the required number of switches, the number of vulnerable trunk lines, and simplifies maintenance. How much easier it is to look for a problem in one switch than in several scattered rooms, as the author of the above article suggests. It is not for nothing that the phrase βswitch zooβ is used.
But often in practice one still has to use either the "fractal star" topology or the "mixed topology". This is due to the limitation of the distance from the switching equipment to the workstation. That is why I believe that optical networks will eventually completely replace twisted pair.
If it is not possible to place all the switches in one place, then it is preferable to use a mixed topology, because. all trunks will go on different routes, which will minimize the likelihood of simultaneous damage to several trunks.
Speaking of trunks. Switches connected by trunk lines should always have a backup channel, then in case of damage to one line, the connection between the nodes will remain and not a single connection will be broken. It will be possible to slowly pull the damaged wire. Therefore, for trunks, even at short distances, you can use a faster and thinner optical patch cord.
The second principle of building SCS is rationality and practicality. It is rationality that does not allow the use of "modern" optics in connecting workstations and other network devices. As the author of the above article correctly noted, everything now works over a twisted pair cable. It's very practical. But on optical channels without additional devices, there is still little that can work. And each additional device is not only a vulnerability, but also an additional cost. But still, this is the future. Someday, when almost every device will have a built-in optical port, optics will completely replace twisted pair.
Rationality and practicality can also be seen in the number of rj45 outlets in the workplace. It is practical to use 2 sockets per place. The second line can be used, for example, to connect an analog (digital) telephone, or simply be a backup. This is how SCS is usually designed for large companies. For small and medium businesses, it is more rational to use one computer socket per workplace, since ip phones basically have two ports - an incoming link and a second one for connecting a computer through it. For network printers, it is always desirable to design a separate workplace, and locate it as conveniently as possible for all employees using it, for example, in corridors. Deciding what is more important - rationality or practicality should be a person competent in the IT field, since we all know perfectly well what management usually chooses.
There is another important point that I would attribute to rationality and practicality. This is reasonable redundancy. It is more practical to have as many workplaces in the offices as can accommodate employees, and not how many people are currently working there. Here, again, a competent employee who has an idea of ββββthe financial capabilities of the company, and understands that in the case of new Wishlist, he will have to solve the problem of lack of places, should decide.
And of course, the choice of equipment and materials can be attributed to the principle of rationality and practicality. For example, if a company is small and does not have the ability to employ a competent network administrator capable of working with L2 switches, it makes sense to use unmanaged switches, while backup trunks should still be, albeit not active ones. You don't need to save on materials. To use a copper-plated cable instead of a copper twisted pair means that in a couple of years they are guaranteed to face the problem of bad connections. Refusing patch panels, factory patch cords and organizers means getting mess in the closet after some time, constantly βfalling offβ links and oxidation of connectors. Don't skimp on a server rack either. The large size not only allows you to place more equipment in it, but also makes it easier to maintain.
Don't skimp on patch cords. Good factory patch cords should be both in the workplace and in the server cabinet. If you calculate the time spent on crimping the connectors and the cost of materials, then buying a factory patch cord will be cheaper. In addition, the cable will be tight, the connectors may be bad, the connectors will oxidize much faster, the crimping tool may be bad, the eye may blur and there are many more reasons not to use a homemade patch cord.
In my opinion, if there is no need for a workstation to operate at 10G speeds, it is more rational to use category 5e twisted pair cable and not category 6, because it is not only cheaper, but also thinner, more flexible, and therefore more convenient to install.
And finally, the third principle is orderliness. The larger the network, the more important the order in it. Sockets and ports of patch panels must be numbered. Numbering usually starts at the workplaces from left to right from the entrance to the premises. There must be an approved floor plan with the location and numbering of outlets.
It is for orderliness and not for the physical separation of networks that patch panels are used. If the author of the βmore than once already mentionedβ article assumes that there is nothing special to switch in his closet, then we cannot afford this.
That's all. These three basic principles guide any of my SCS projects. In this article, I could not touch everything, I probably missed a lot, somewhere I might be wrong. I am always ready for a constructive discussion if I am given an invitation or in personal correspondence.
Source: habr.com